Legacy implementation made incorrect assumptions on how many bits
of useful information is returned by libc's `rand()`. If `RAND_MAX`
had more than 16 useful bits, it assumed that `rand()` could return
32 useful random bits. However, e.g., jerry-libc's `RAND_MAX` is
`0x7fffffff`, which denotes 31 useful bits only. The consequence
was that `Math.random()` covered only the lower half of the
standard-mandated `[0,1)` interval.
This path fixes the error and always uses the exact value of
`RAND_MAX` to compute the random value, which will thus fully cover
`[0,1)`.
Fixes#1414
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
`-g` is enough (most modern compilers use DWARF4 as default, at
least, e.g., GCC has switched from DWARF2 in 4.8), more future
proof (if a compiler moves to a newer version, it wont get forced
to a fixed older version), and less restrictive (some compilers
handle the "standard" `-g` only but don't understand its variants).
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Defining the same macros in multiple source files can cause macro
re-definition warnings or errors in all-in-one builds.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Newer compilers (especially clang) warn (and fail) on
`assert (!"message");` constructs typically used to assert on
unreachable code paths (multiple occurrences in jerry-libc). A more
up-to-date approach is to use `assert (false && "message");`. This
patch applies the pattern to all relevant asserts in jerry-libc.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
This patch improves the command line tool in two ways:
* The tool can be instructed to read and execute a script from
stdin by invoking it as `jerry -`.
* The tool can be instructed to suppress its prompt when in repl
mode by invoking it as `jerry --no-prompt`.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Several non-script files have been added with exec permission to
the repository. This patch revokes the erroneous permissions.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
run-tests.py used to echo the build commands it executed, but did
not behave the same way when it executed test commands. To make its
behavior more traceable, this patch adds echo for test commands as
well.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
From the beginning, we have been configuring cppcheck to check its
input as C++ source. However, the transition to C99 happened a
while ago. This patch switches cppcheck into C99 mode.
Some related changes:
* Progress reporting of cppcheck just clutters the output and makes
warnings hard to discover. Thus, this patch puts cppcheck into a
quieter mode where it prints anything only if a non-suppressed
warning is found.
* The default warning format of cppcheck is a bit different from
usual compiler error/warning format. This patch configures
cppcheck to use a more familiar warning template.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
It's stated in the Guidelines that all contributions must be under
the Apache License 2.0. To avoid potential mistakes from manual
reviews, this patch adds the check-license.py script to
automatically check all source files for license headers.
Travis CI is also configured to run the check.
Fallout: it turned out that some files already in the code base
either miss a license header or have some minor typo differences.
The patch fixes up some of these deficiences.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The naming convention of the project for `struct`s is
`typedef struct x_t { } x_t`, but only if it has self-recursive
pointer members, otherwise `typedef struct { } x_t` is enough.
This patch applies this style consistently throughout the code
base.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* Cmake had LTO OFF, build.py had LTO ON by default. ON became the
common setting.
* Cmake had snapshot save/exec OFF, build.py had them ON by
default. OFF became the common setting.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Although both jerry-libc and jerry-libm have configuration options
that enable/disable their build, in practice, only jerry-libc can be
replaced with the system (compiler-default) libc. If jerry-libm is
disabled, the build of jerry-main fails, as there is no way to
instruct the linker to link the system libm to the binary. (The
build system does have a way to pass flags to the linker, but those
flags are listed before the linked objects. For the references to
get resolved correctly, the libraries to be linked have to be
specified _after_ the objects.)
This patch adds the EXTERNAL_LINK_LIBS configuration option to
CMakeLists, which ensures that the specified libraries get
correctly passed to the linker. (E.g, replacing jerry-libm with
system libm becomes possible with
`JERRY_LIBM=OFF EXTERNAL_LINK_LIBS='-lm'`.)
Additionally, the patch also makes the following related changes:
* Removes the COMPILER_DEFAULT_LIBC configuration option, as it is
(almost) always the opposite of JERRY_LIBC. Moreover, its name is
misleading: its only role is to add `-nostdlib` to the linker
flags.
* Makes use of transitive library dependencies: if a library has
another library as dependency, and it is linked to a binary, its
dependency is linked as well. Thus, jerry-libc, jerry-libm, and
any external libraries are added to jerry-core as dependency, and
then only jerry-core is linked to executables (cmake will take
care of the rest).
* build.py and run-tests.py follow up the changes, along with some
minor syntax changes.
* Moves static linking option to global CMakeLists, as unit test
binaries should be linked the same way as jerry-main.
* Adds EXTERNAL_COMPILER_FLAGS and EXTERNAL_LINKER_FLAGS as last to
the flag list, to allow user override of (nearly) anything.
The patch speculatively follows up the build system changes in the
mbed, riot-stm32f4, and zephyr targets.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The new build script does not support ninja anymore, and we haven't
been building bare-metal arm targets for long. This patch removes
now-obsolete dependencies from install scripts.
This can also help to reduce the load on Travis CI (as it keeps
installing dependencies for each build job over and over again).
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
With the new build system, the conventional `cmake && make` already
works. However, the last step, i.e., `make install` was still
missing (didn't work). This patch adds the `install()` commands to
CMakeLists that are required to generate the `install` target in
the Makefile.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
`tools/build.py --help` prints a lot of output, which is sometimes
hard to comprehend, especially since some described options are
useful for developers only.
This patch:
* shortens how options are displayed in the output with the help of
the "metavar" feature of the argparse module ("--option {on, off}"
becomes "--option X" and the choices are listed in the description
only); and
* hides all developer-specific options from "--help", and reveals
them only for "--devhelp".
As a result, help output becomes more readable, especially for
ordinary users, and needs less space.
Some additional changes:
* the options are ordered in a more logical way (although that's
somewhat subjective);
* help strings start with lowercase letters (it aligns better with
argparse's overall style); and
* rename "--cpointer_32_bit" to "--cpointer-32bit" (underscores are
alien to command line options).
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Until now,
* the link map of jerry-main has always been generated, even though
it is not used that often;
* static linking has always been forced except for the OSX target.
This patch adds configuration options for these features.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
In tolerant mode, only the existence of the Signed-off-by line is
checked but the name and address is not matched against the author
of the commit.
The tools/run-tests.py script is extended to allow calling the
check script in tolerant mode and also to detect whether it is
being run by Travis CI.
The Travis CI config has been modified to check PRs in strict mode
but use tolerant mode for merges to master. This should enable the
use of the Merge button on the GitHub web interface. (The PR is
strictly checked when a contributor opens it but when a committer
merges it via the web interface and GitHub rewrites author details
from the contributor's profile, master will not turn red either.)
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Previously, jerry-core was built with various build identification
info (build date, git commit hash, git branch name), which was both
available from string constants of the library and also exposed by
jerry-main via the `--version` command line option. As of late,
jerry-core identifies itself only with API version number macros,
but sometimes it could be helpful to get access to the git commit
hash as well, even if that info does not become part of the API.
This patch moves the git commit hash retrieval logic from the build
scripts of jerry-core to jerry-main, and changes `--version` handler
to print the hash as well.
Resolves#1295
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
It should not import things it does not use and it should define
settings data only (i.e., definition of functions should not happen
there).
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
OS X build regularly reports some 39 warnings falling in 3 major
categories:
* "static function '...' is used in an inline function with
external linkage [-Wstatic-in-inline]": Some semantics around
`inline` have changed between C89 and C99, and gcc and clang seem
to disagree on how strict they should be about them. Solution
chosen is to use `-Wnostatic-in-inline` command line option for
clang.
* "implicit conversion turns floating-point number into integer:
'double' to 'bool' [-Wfloat-conversion]": `if (fmod (..., ...))`
was used at different places, which is not nice anyway, thus the
return value is compared explicitly against `ECMA_NUMBER_ZERO`.
* "format string is not a string literal [-Wformat-nonliteral]":
Console and log port I/O functions have a printf-like interface,
and the default implementations actually pass both format string
and the remaining arguments to a vfprintf. However, clang is
strict about the format string parameter of vfprintf and expects
a literal there. By annotating the port I/O functions with
`__attribute__ ((format (printf, ..., ...)))`, clang will check
the format string being a literal string earlier, when the port
functions are called, and will not complain within them when
vfprintf is called.
(Actually, this has revealed an incorrect format string, which
has been fixed as well.)
(There were also some single conversion errors not listed above.)
The patch was tested on OS X (where all warnings disappeared), but
it should help clang compilation on other OS's as well.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* Split list into two: what is needed for building and what is
needed for development only.
* Remove g++ from install list.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* Remove JERRY_CORE CMake option: the building of the core
JerryScript library should not be optional.
* Fix wording of comments, status and error messages.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
For 64 bit pointers, 0xffffffff as end-of-list marker pointer does
not work, as the marker is expected to be greater than the start
address of the heap area (and also max 0xffffffff away), but on 64
bit systems, the heap can start on higher addresses, i.e., above
0x100000000.
This patch changes JMEM_HEAP_END_OF_LIST from pointer to offset.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Apple does not support staticaly built applications on OSX, and all
dynamically built apps have to be linked against the System lib
(i.e., `-static` should not be used and `-lSystem` is a must).
As System contains all libc and libm functions, building (and
linking) the minimal jerry-libc and jerry-libm libs makes no sense.
Moreover, if JERRY_LIBC is ON, the compiler will use the jerry-libc
headers but will link the libc functions from System, which causes
heavy confusion and segfaults at run time.
Thus, this patch changes the build system to disable the building
of jerry-libc and jerry-libm, and enables the use of system
libraries when building on OSX.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Remove "-I/home/.../jerryscript/jerry-libc/__TARGET_HOST" and
"-D__TARGET_HOST" from compiler command line, they don't add
anything useful.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The Math.pow implementation relies on libm's pow. However, the ISO C
and ES5.1 standards differ on pow:
* `x ** NAN` is NAN in ES but `+1 ** y` is 1 in C
* `+-1 ** +-INF` is NAN in ES but 1 in C
This patch:
* Modifies the Math.pow implementation to handle the special cases
instead calling pow.
* Adds a test case to jerry-test-suite as it did not test
`Math.pow(1,NaN)`.
* Fixes jerry-libm's pow, as it was not standard conforming, which
helped hiding the error in Math.pow.
* Updates the unit test for libm.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The manual of `mktemp` states that "TEMPLATE must contain at least
3 consecutive 'X's in last component." Linux implementation seems
to be relaxed about this and accepts and rewrites X's even inside
the template, but mktemp of OSX is more strict and handles trailing
X's only. This patch makes sure that mktemp templates work on both
OS's.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
This patch:
* Ensures that all calls to `jerry_port_log` in jerry-core happen
via macros defined in jrt.h. Also, it unifies the names of those
macros: as `JERRY_ERROR_MSG` and `JERRY_WARNING_MSG` gave a good
pattern that was well aligned with the naming scheme of the log
level enum, `JERRY_DLOG` and `JERRY_DDLOG` were rewritten to
`JERRY_DEBUG_MSG` and `JERRY_TRACE_MSG`.
* Ensures that all debug logging code parts of jerry-core (i.e.,
memory statistics, JS byte-code dumps, and RegExp byte-code
dumps) are guarded by macros: `JMEM_STATS`,
`PARSER_DUMP_BYTE_CODE`, and `REGEXP_DUMP_BYTE_CODE`, which in
turn are controled by cmake build system feature flags
`FEATURE_MEM_STATS`, `FEATURE_PARSER_DUMP`, and
`FEATURE_REGEXP_DUMP`.
* Ensures that all debug logging functionalities can be controled
during run time (provided that they were enabled during build
time): the engine has `JERRY_INIT_MEM_STATS[_SEPARATE]`,
`JERRY_INIT_SHOW_OPCODES`, `JERRY_INIT_SHOW_REGEXP_OPCODES` init
flags, and the default unix/linux command line app has
corresponding command line switches.`
* Drops `FEATURE_LOG`, `JERRY_ENABLE_LOG`, and
`JERRY_INIT_ENABLE_LOG`, as their name was misleadingly general,
even though they mostly controled the regexp engine only. The
above-mentioned `*REGEXP*` things mostly act as their
replacements.
* Updates build, test, and measurement tool scripts, and
documentation.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The `cmake` directory already contains several toolchain files for
various platforms (operating system + architecture). However,
`tools/build.py` does not define a toolchain file for cmake unless
explicitly specified. This patch changes the script to look into
the `cmake` directory for a file named
`toolchain_$(os)_$(arch).cmake` and, if found, pass that to cmake
by default.
OS and arch are determined by `os.uname()`. As Linux on Raspberry
Pi identifies itself as "armv7l", the legacy "armv7l-hf" arch name
is shortened to "armv7l". This way, building jerry on RPi
(natively, not cross) becomes possible by simply running
`tools/build.py` without any extra options.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The fact that the JS parser had been developed as a separate
component for a while is still visible from some macros that mirror
things from JRT. This patch removes those duplicates and makes the
JS parser rely on jrt.h. (The removed macros are: `PARSER_DEBUG`,
`PARSER_INLINE`, `PARSER_NOINLINE`.)
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
CMakeLists already contains macros to ease adding compilation and
warning flags. This patch:
* Ensures that they are used whereever possible.
* Adds more macros to help with other flags as well.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
`jrt_read_from_buffer_by_offset` is not used anywhere in the code
while `jrt_write_to_buffer_by_offset` is only used by snapshot
saving functions. Thus, this patch removes the read variant
completely and moves the write variant as a static function to
jerry.c. This empties out jrt.c, thus deleting.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Until now, jerry had 3 different assert-like routines:
`jerry_assert_fail`, `jerry_unreachable`, and `jerry_unimplemented`,
and 3 corresponding macros (`JERRY_ASSERT`, `JERRY_UNREACHABLE`,
and `JERRY_UNIMPLEMENTED`). They had some irregularities, namely:
* All of them had a string parameter, although `jerry_unreachable`
never got anything there but NULL.
* Both `jerry_unreachable` and `jerry_unimplemented` checked its
string parameter for NULL, although it was always NULL for the
first one and never NULL for the second.
* `jerry_unreachable` is just a regular assert with a fixed error
message (i.e., control should not have got here), however, the
expansion of its corresponding macro in debug and release modes
differs from the behaviour of `JERRY_ASSERT`: `JERRY_ASSERT` is
a no-op in release, however, `JERRY_UNREACHABLE` was triggering
a crash even there.
* Moreover, `JERRY_UNIMPLEMENTED` was almost never used anymore but
in a few places (where often an `#ifdef` selected between
`JERRY_UNIMPLEMENTED` and `JERRY_UNREACHABLE`).
Because of the above, this patch makes the following changes:
* Drops `JERRY_UNIMPLEMENTED` completely and whereever it was still
used, replaces it with `JERRY_UNREACHABLE`. As a consequence, the
`jerry_unimplemented` function and the `ERR_UNIMPLEMENTED_CASE`
fatal error code are also removed.
* Makes `JERRY_UNREACHABLE` expand to no-op in release builds.
(Actually, to `__builtin_unreachable ()` to avoid warnings.) As
a consequence, makes both `jerry_assert_fail` and
`jerry_unreachable` be guarded by `#ifndef JERRY_NDEBUG`. Also,
changes `jerry_unreachable` not to expect a string parameter.
* Rewrites `TEST_ASSERT` not to rely on `jerry_assert_fail` as
`TEST_ASSERT` has to work in release builds as well. This also
allows changing the error message not to mention "ICE", which
would misleadingly suggest an assert within the engine, but
"TEST" instead.
As a side-effect of the cleanup, some refactorings happened in
jrt.h:
* Removed the definition of the unnecessary `__extension__` macro.
* Re-used `JERRY_UNUSED` and `unlikely` where possible.
* Moved some parts of the file around.
* Fixed some comments (`/**` should only be used for the docstring
of a single entity, for groups header comments, the regular `/*`
should be used).
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
As the comment of syscall error mentions, we should not try to log
(or print) anything in that case.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
There is no need for putting `LIT_CHAR_*` constants in const
scalars or arrays, they can be used directly. (There is especially
no need for arrays of 10 elements if only 2 of them are used.)
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* Make constants static in `ecma_number_to_utf8_string`.
* Make `ecma_number_to_utf8_string` use early returns, and rewrite
its self-recursion in case of negative numbers.
* Make the stringification of decimal exponent in
`ecma_number_to_utf8_string` use `ecma_uint32_to_utf8_string`.
* Changed ERROL0 dtoa implementation to use the `double` type
instead of `ecma_number_t`. Thus, even is `ecma_number_t` is 32
bit wide, the algorithm works the same.
* Changed `ecma_number_to_decimal` to use the ERROL0 dtoa algorithm
for 32-bit floats as well.
* Changed `ecma_number_to_decimal` to generate the decimal string
representation of the mantissa instead of an `uint64_t` number.
* Changed `ecma_number_to_utf8_string` to make use of the already
available string representation of the mantissa, generated now by
`ecma_number_to_decimal`.
* Changed `ecma_number_to_utf8_string` not to use static arrays and
variables for digit, "e", etc. generation.
* Changed all `Number.prototype.toXXX` implementations and the
`ecma_builtin_number_prototype_helper_round` helper to make use
of the already available string representation of the mantissa,
generated now by `ecma_number_to_decimal`.
* Factored out the common stringification parts of all
`Number.prototype.toXXX` implementations into a new helper
`ecma_builtin_number_prototype_helper_to_string`.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* Improve the name of the temporary snapshot files in test runner:
Until now, test runner used a completely random filename for the
temporary snapshot file when testing (saving to and executing from)
snapshots. This patch makes the temp file contain part of the name
of the original test file in order to make reading the logs (and
thus, identifying failing tests) easier.
* Add `--snapshot` to summary info of test runner if testing snapshot
support:
This helps reading the logs, especially when multiple tests run in
parallel and get their output mixed. E.g., on the CI, all tests
write the console at the same time and messages get interleaved.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
When running multiple test suites in parallel -- e.g., as it
happens on the CI --, the names of the executed commands and the
results of the executions can/do get printed far from each other,
interrupted by other prints. This can make the reading of the
output hard. This patch prints each line with one echo command,
which makes the interruption of the line less likely.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Jerry (the command line tool) has been supporting the execution of
multiple script files for long. However, until now, it simply
concatenated all sources into a single source buffer and
parsed/executed them as one unit. Other JS execution tools (e.g.,
jsc, v8) load and execute separate files as separate units -- but
still in the same execution environment. The most significant
effect of this approach is that the `"use strict;"` directive (or
the absence of it) at the beginning of each JS script file takes
effect as expected (i.e., as if the script was executed alone).
Contrarily, the concatenation-based approach forces the strictness
of the first script on all the rest (i.e., if the first script
starts with `"use strict";` the rest is also executed in a strict
environment even if they did not contain the directive, and vice
versa).
This patch makes the jerry command line tool to load/parse/run one
unit at a time.
Side effects:
- As there is no need for separate file read routines that load one
file (a snapshot) or concat multiple (JS sources) anymore, those
routines got merged.
- Both previous read routines used multiple stdio functions
(`fseek`, `ftell`, and `rewind`). This has been simplified to
rely on `fread` only to find out the length of the input.
- This simplification made the above mentioned functions
superfluous in jerry-libc.
- As some error messages had to be touched in this patch, several
more have been beautified to make them more consistent.
- One small change was needed in `jerry_parse` in jerry-core to
allow subsequent parsing of multiple sources (without that, an
assertion was triggered).
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Both JS test suite and unit test logs can grow large as Makefile
invokes run-test-suite.sh and run-unittests.sh with absolute paths
to engines and test directories, which get then printed quite
often. This patch adds code to the runner scripts to determine the
longest directory path common to the current working directory, the
invoked script, the test directory, and the engine (in case of JS
tests). Then, when a path is to be printed, this common path
component is skipped.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
When merging PRs via the GitHub web interface, CR characters can
sporadically appear in the commit message. This can also cause
signed-off-by message mismatch in check-signed-off.sh (even if
everything else is OK). To be on the safe side, this patch removes
any CR characters from the output of git show when performing the
check.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Recent changes eliminate the need for `jerry_port_putchar`. As port
API discussions don't make it likely that it will ever be needed
again, this patch removes its declaration from jerry-port.h and its
implementations from the port(s).
The related code in jerry-libc is not needed either: whatever `putc`
(and `puts`) can do, `printf` can do as well.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The internals have surprisingly large number of unused functions at
surprising places. This patch cleans them out to make maintenance
easier and to prevent dead code having effect on future decisions.
The patch intentionally does not try to clean up public API but
focuses on internal modules only that are/should not be reachable
from "outside".
However, unit tests do access private API, thus tests of literal
storage had to be adjusted.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Replaced `gettimeofday`-related code with `jerry_port_get_current_time`
and `jerry_port_get_time_zone` function calls. Moved old code to
default port implementation.
Removed `ENABLE_DATE_SYS_CALLS` as date syscalls should "just work".
Fix DST adjustments: even if `gettimeofday` returns meaningful data in
`tz_dsttime`, the value is just a flag (or, according to some sources,
a tri-state value: >0 if DST applies, ==0 if DST does not apply, <0 if
unknown). Hitherto, the field was simply added to/subtracted from a
time value in milliseconds. To use it approximately correctly, the
field's value should be multiplied by "millisecs/hour".
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* Moved the error codes to jerry-port.h and declared port function
`jerry_port_fatal`.
* Moved "exit or abort on fail" functionality to the newly added
jerry-port-default-fatal.c.
* This implied that a default port-specific API had to be introduced:
functions `jerry_port_default_set_abort_on_fail` and
`jerry_port_default_is_abort_on_fail` declared in jerry-port-default.h
control the fatal exit behaviour.
* For the sake of clarity, renamed jerry-port.c to
jerry-port-default-io.c.
* Adapted CMakeLists to deal with port implementations consisting of
more then one source file and exposing headers. This also required
the renaming of `EXTERNAL_PORT_FILE` cmake option to
`EXTERNAL_PORT_DIR`.
* Adapted main sources to use the default port header for the
abort-on-fail functionality, as that is not part of the core jerry
API anymore.
* Added default port implementation to the static source code checker
tools.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* LICENSE needed year update.
* The asm component of the posix target of jerry-libc had no
copyright & license notice since it has been introduced in 2015.
Traced back history and added missing header with correct years.
* Three tests in jerry/fail/1 also missed header. Added.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Apps that use jerry should not (must not!) include and rely on its
internal headers. Typical issue is the use of "jrt/jrt.h". Fixing
`main-{unix,mcu}.c`, and speculatively the apps under the `targets`
directory as well.
(Note: a fix can be either including "jerry-port.h" and using
functions declared there, e.g., `jerry_port_errormsg`, or simply
using standard libc function like `printf`. Both approaches occur
in this patch.)
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The two target implementations are very close clones of each other.
The only known differences are the following:
* The asm component of the linux target has `.type` and `.size`
declarations for the functions defined therein, while the darwin
target doesn't support those.
* Linux uses `__NR_xxx` mnemonics for syscall numbers, while darwin
uses `SYS_xxx` format.
* Darwin does not have `exit_group` syscall but `exit` only.
* The linux target has a commented out `jrt_set_mem_limits` fuction
declaration at the end of the C source. (Based on its name, this
function does not really belong here.)
Simple preprocessor macros can unify the first three differences.
While for the sake of legacy, we can keep the fourth commented-out
code in the code base; it might turn out to be useful elsewhere in
the future. Since it remains commented out it wont cause any
problems on any OSs.
So, this patch gets rid of a lot of duplication.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Introducing the `assert.h` public header with the `assert(x)` macro
and already making use of it in jerry-libc sources. These newly
introduced ISO C-conforming features replace the non-conforming
`LIBC_ASSERT` and `LIBC_NDEBUG` at no cost, and they also have the
benefit that we can expose `assert` in a public header and make it
useful to jerry-libc users.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Fixed doc comments issues:
* Fixed mistyped param doc comments (`/**<` is OK, `/** <` is not).
* Put special characters (e.g., pipe, backslash, etc.) in quotes, as they can
confuse doxygen and it will print lots of various warnings. For the sake of
completeness and consistent style, also quote some special characters in
re-bytecode.h
* Added missing `@{`s, removed extra `@}`s.
* Turned `/*` comments to `/**<` doc comments.
Ensured same style for doc groups everywhere:
* Where `\addtogroup`, `@{`, and `@}` doxygen commands are used, the order to be
followed is: license, `#ifndef` guards (in headers), includes, `\addtogroup`
and `@{`, main code content, `@}`, `#endif` guards (in headers).
* Multiple `\addtogroup`s or multiple `@}`s should be in the same doc comment.
* First `\addtogroup` should be on the very first line of a doc comment, i.e.,
`/** \addtogroup`.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The core functionality (i.e., the equivalent of `kill (getpid (), sig);`)
was already there in the implementation of `abort ()`. Now, it got
factored out to `raise ()` so that others (most importantly,
`__aeabi_ldiv0`) can call and link to it as well.
Also, removed `LIBC_UNREACHABLE_STUB_FOR` macro, as it is not used anymore.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
After recent changes to jerry-libc, `syscall_N` C functions became
pure trampolines to their appropriate `syscall_N_asm` counterparts
written in assembly. Removing the C functions and renaming the
assembly functions to take their place simplifies the code.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Discussions on Termination Port API point in a different direction
than what's already in the API. Moreover, since the function has
been throwing unimplemented assert error since its introduction,
we should remove it from the code base.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* Rename modified fdlibm to jerry-libm
* Move third-party/fdlibm to jerry-libm
* Rename original fdlibm.h to jerry-libm-internal.h
* And remove it from the public headers.
* Rename jerry-libm's public header to math.h
* This also makes jerry-core sources include `<math.h>`. Therefore,
should anyone want to use a different libm implementation with
jerry, it becomes possible. (The same way as we provide a minimal
libc with standard headers, but should it be insufficient or
conflicting for someone, it can be replaced.)
* Drop `s_` prefix from jerry-libm sources
* The original fdlibm implementation had various prefixes (e.g., `k_`
for sources of kernel routines, and `w_` for wrapper routines), but
after the specialization of fdlibm to jerry, only `s_` remained.
Since it does not encode anything anymore, it can be dropped.
* Stylistic edits to jerry-libm's CMakeLists
* Align project name with other CMakeLists in the code base
* Move Jerry-LibM under Apache License
* Using the same approach as was used by linux-wireless when ath5k
driver license needed clarification. Solution was proposed by SFLC.
External mail for future reference: http://lwn.net/Articles/247806/
* Tests & checks
* Remove FD from the name of libm unit test-related files
* Make vera++ and cppcheck check jerry-libm
* Targets
* Speculative update of targets to use jerry-libm
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Endianness was not properly set/detected for AArch64. Once that is
fixed, the platform is well-supported.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Those macros are legacy and are not used consitently throughout the
code base. This patch eliminates their definitions and rewrites
their remaining occurrences to TODO comments.
All occurrences have been checked and made sure that the comments
used a consistent style.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Re-styling of the sources revealed that the false branch of an
if-else construct in `asin` did not contain all the statements
that it should. (This was an issue -- a bad smell at least --
that's been around since 1995, according to the legacy SCCS ID.)
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* First re-style was done automatically by indent to minimize the
chance of errors during rewrite.
* Manual changes were applied to non-critical places only (comments
and spaces):
* Replaced all tabs with spaces.
* Fixed tab stops in formulae in function comments.
(Note: ASCII art for math formulae (especially for super- and
subscripts) is a terrible idea.)
* Unified the style of function comments.
* Moved some in-code comments to their right places, which indent
couldn't handle.
* Added spaces to formulae of in-code comments to make them more
readable.
* Added braces mandated by jerry style guidelines.
* Added parentheses to multiline #ifdef.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* Made helper functions static.
* Changed the signature of __kernel_rem_pio2 not to require the
ipio2 argument
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Keep IEEE code paths only:
* removed SVID, XOPEN, POSIX code paths from everywhere.
* deleted s_lib_version.c, as version is only useful if multiple
standards are supported.
* deleted k_standard.c, as it handles non-IEEE exception cases only.
* renamed the e_{acos,asin,atan2,exp,fmod,log,pow,sqrt}.c sources as
s_.*, dropped the __ieee754_ prefix from the names of the
appropriate functions therein, and deleted the
w_{acos,asin,atan2,exp,fmod,log,pow,sqrt}.c wrapper code.
Keep C99 declaration variants only:
* removed old C-style function declaration variants.
* removed data declaration variants where const qualifier was not
used.
Clean unused sources/functions:
* removed s_{rint,significand,tanh}.c and the appropriate functions
defined therein.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
MEM_HEAP_PTR_64 is duplicating existing information: stdint.h, which
header is already used by the project, defines various _MAX macros
for upper limits of integer types. The comparison of UINTPTR_MAX and
UINT32_MAX can give the same info as encoded in MEM_HEAP_PTR_64.
The stdint.h-based approach has the benefit that jerry can support
any 64-bit architecture without the need for editing the build
system. (With the existing approach, CMakeLists has to know about
every 64-bit architecture to work properly.)
Thus, removing the extraneous macro from the code.
The patch also changes the mem_pools_chunk_t struct, as it turned
out that the struct does not have to be padded to MEM_POOL_CHUNK_SIZE.
(The padding also depended on MEM_HEAP_PTR_64.) It is enough if the
size of the struct is smaller than (or equal to) MEM_POOL_CHUNK_SIZE.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The project is relying on a variant of fdlibm, which has aleady
been edited but never verified for correctness. This patch adds
unit testing of fdlibm by:
* introducing a test generator that uses a trusted libm
implementation to calculate correct and expected results of math
functions
(tools/gen-test-fdlibm.sh and tools/unit-tests/gen-test-fdlibm.c),
* adding tests created with the generator that stress all publicly
exported functions of jerry's fdlibm
(tests/unit/test-fdlibm.inc.h), and
* adding a unit test file to drive the generated tests
(tests/unit/test-fdlibm.c).
Note: The test generator is not expected to be executed often, thus
it is not wired into the build system. If it gets edited, it must
be used locally to re-generate the .inc.h file.
During development, it turned out that tests/unit/test-common.h
included the system header math.h, which was only a bad smell until
now but became a real header conflict issue with the introduction
of the fdlibm unit test. Thus, this patch also changes the include
to fdlibm-math.h.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Currently, if a regexp contains a character class and that
character class contains a hex-specified code point -- either
\uNNNN or \xNN --, then the character 'u' or 'x' is added to the
character class as well. This patch fixes the error and also adds
a regression test covering the issue.
Fixes#962
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
In case of literals and in main, snapshot saving is incorrectly
named dumping. Elsewhere in the code, 'dump' functions output debug
data (even literals have a `lit_dump_literals` debug function). To
help distinction and to align terminologies, snapshot saving
functions of literals and command line options of main are also
re-named 'save'.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Most of the legacy version control IDs in fdlibm are already simple
comments at the beginning of the source files. However, in some
files, there are leftover global variables and pragmas. (The global
variable in e_pow.c even increases the .data section in both debug
and release builds.) This patch turns all legacy IDs to comments of
unified style.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Make the internal heap and pools memory usage statistics APIs more
similar: how the print functions are named, where they are
implemented, and which parts of them are guarded by `MEM_STATS`.
Also, adapt unit tests to the changes.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
* Guard `lit_dump_literals` with `JERRY_ENABLE_LOG` (both in source
and in header).
* Change `printf`s to `JERRY_DLOG`.
* Make `lit_dump_literals` be called in `lit_finalize` (it was dead
code).
* Remove its duplicate declaration from lit-literal.h
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The new CBC interpreter does not support it anymore, thus removing
related code.
(As a side-effect, `jerry_flag_t` has been refactored from
`uint32_t` and associated defines to an enum.)
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
There have been several renamings of header files but the
ifndef/define/endif preprocessor directives did not always follow
the file name changes (or perhaps never followed a unified naming
scheme). This patch gets all headers aligned with the style of the
majority.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Cppcheck has already been used for checking unit test sources
statically but vera++ style-checking of unit tests was left out,
most probably only by mistake. This patch adds unit tests to the
set of style-checked sources and also fixes the stlye issues that
were present in the test code.
(Extra: since the project is pure C now, not trying to collect
*.cpp files anymore.)
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
E.g.,
* `ssize_t` was used where `lit_utf8_size_t` or `jerry_api_size_t`
would have been correct,
* `lit_utf8_size_t` was used where `ecma_length_t` would have been
correct.
Note, the patch also includes internal and public API changes:
* `ecma_string_to_utf8_string` does not return negative value if
output buffer is not large enough to contain the string; the
buffer is expected to be large enough. (`ecma_string_get_size`
can be used to retrieve the required size.)
* `jerry_api_string_to_char_buffer` adapts the same logic (and
`jerry_api_get_string_size` can be used to determine the
required size of the buffer).
Related issue: #942
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
There was an abuse of memory pools in jerry.c: `compiled_code_map_entry_t`
is not an ECMA data model object but was still alocated/freed via
mem pool functions. Changed the appropriate calls to heap-related
functions.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
It is already supported by CMakeLists, might be useful to allow it
to be passed through Makefile.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The repeated downloads of the STM packages cause a huge slow-down
of Travis, especially when the source site becomes irresponsive
(which even causes Travis to report errors when the build would
work fine otherwise). Unfortunately, the license of the STM
packages does not allow their inclusion in the repository. (Note:
locally executed `make precommit` still builds MCU targets.)
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
The handling of option defaults is already done in CMakeLists, it
is no good practice to duplicate the logic in the Makefile.
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
They have been kept when the build system was refactored but it
turns out that they are not used at all (they are not passed on to
cmake).
JerryScript-DCO-1.0-Signed-off-by: Akos Kiss [email protected]
Akos Kiss